Re: On autogenerated ChangeLog

[Garrett LeSage <garrett novell com>]

On 04/21/2009 10:23 PM, Zeeshan Ali (Khattak) wrote:
> 1. All information in the ChangeLog is redundant.
> 2.  Maintaining a ChangeLog only and only realizes otherwise
> inexistent conflicts.
>    

Agreed.

You could do what we've been doing in a project of mine:

Auto-generating the changelogs on release, and publishing a hand-picked 
"friendly" changelog too.

In fact, you probably should have a file which would be updated as 
developers churn along (distributing the effort over time), where each 
big notable feature/bugfix/etc. is recorded as a line in the file 
whenever it officially lands. When you make a release, create a tag in 
git (using "git tag -a" of course). When you diff the file between tags 
(this can be done easily if you have a "release" branch which always 
points at the most current release) and extract only the new lines, you 
have a changelog of well-written, user-friendly features — perfect for 
posting to your project's website, sending out to a mailing list, etc.

Then, for those who want a more traditional "here's-everything" 
changelog, just generate a changelog based on your git commit messages. 
This should also be done also automatically on release.

The nice thing about a process like this is that it minimizes the effort 
needed (since it's all done automatically using git, at release time) 
and should make everyone happy.

Garrett